One hundred ninety patients with chronic angina for an average of 2 years were followed. Forty-seven had at least 1 mm of ST deviation in response to provocation of coronary vasoconstriction by prolonged hyperventilation (group I); 143 had no ST deviation (group II). The angiographic response to this test was studied in 21 patients from group I, and revealed 25 to 100% diameter reduction; in group II 9 patients showed a 5 to 14% diameter reduction. In group I, 15 patients (32%) died (hazard rate = 0.17 deaths/patients X years) vs 18 (13%) in group II (hazard rate = 0.06) (p less than 0.01). Seven patients in group I (15%) and 3 in group 2 (2%) died while waiting for surgery (p less than 0.01). All patients who died had coronary stenoses of at least 70%. A Cox regression analysis, using 24 variables (invasive and noninvasive), showed a positive hyperventilation test (ST deviation at least 1 mm), low ejection fraction and systolic blood pressure of at least 160 mm Hg to be independent predictors of death (p less than 0.05). Considering only deaths in non-operated patients (patients waiting for surgery and patients not planned to undergo operation), a rate-pressure product/100 of 150 or less at exercise stress testing, left ventricular end-diastolic pressure of 15 mm Hg or more and duration of angina less than 1 year were also independent predictors of death. Thus, the hyperventilation test may be useful for identifying angina patients who are at high risk of cardiac death due to dynamic coronary obstructions.